31st Conference of the
European Working Group on Acoustic Emission (EWGAE) – We.2.A.3
1 License: http://creativecommons.org/licenses/by/3.0/
Localization of Initial Cracks in Laminated
Glass Using Acoustic Emission Analysis –
Part I
Gerd MANTHEI, Christian ALTER, Stefan KOLLING,
THM Technische Hochschule Mittelhessen, Gießen, Germany
Abstract. Laminated glass is experimentally investigated by bending tests with
four-point supported glass panes. The objective of the study is to identify the origin
of initial cracks during testing in order to develop a physically based model for finite
element (FE) simulations of laminated glass under impact load. Tests and FE simu-
lations of windscreens and plane glass plates are carried out and the results are com-
pared in terms of elastic behaviour, maximum load at failure, post breakage behav-
iour, and the fracture pattern. The localization of the initial cracks is based on the
acoustic emission analysis (AE analysis).
1 Introduction
Laminated glass is build of two or more plates of glass together with a polyvinyl butyral
(PVB) interlayer. In this context the interlayer fix the broken splinters and fragments of
glass to avoid injuries of pedestrians and occupants in the case of car accidents. After
cracking, laminated glass is still able to transmit loads because of the interlayer. To simu-
late this effect using finite elements, reliable and physically based material models are
needed in an engineering environment. In an on-going project with focus on pedestrian pro-
tection, head impact tests on windscreens were done and simulated with the explicit solver
LS-DYNA. The objective of the project is to develop a computational modelling technique,
which is capable to predict structural components made from laminated glass under impact
load [1-3]. This includes the behaviour before and after damage as well as an approxima-
tion for crack propagation. In particular the simulation predicts the location of the initial
crack. Thus the windscreens and the flat glass plates are mounted on four pins with a dis-
tance of 1000 mm x 700 mm. An impactor, representing a human head is crushed on the
glass powered by a pneumatic cylinder [4]. The windshields are tested in concave and in
convex position. During the tests the acceleration, the displacement of the impactor, the
deflection of the glass plates, and the force between both are measured. In order to validate
the accuracy of a standard FE model for head impact tests on windshields AE measure-
ments were carried out to localize the initial crack using six AE sensors.
Mor
e In
fo a
t Ope
n A
cces
s D
atab
ase
ww
w.n
dt.n
et/?
id=
1757
3
2
2 Experimental Procedure
2.1 Experimental Set-up
Windshields of cars were tested under static loading conditions. For these tests twelve panes
were available. Left-hand side of Figure 1 presents the testing frame utilized for quasi-static
testing. This consists of a desk-like welded steel construction with pyramidal set-up, which
holds the loading pin including a pneumatic cylinder of SMC, Type C92LA. This cylinder of
100 mm inner diameter can be pneumatically filled with pressurized air up to 10 bar. The
maximal stroke is 400 mm. This cylinder has a special locking unit, which is applied for lock-
ing the piston in arbitrary position. Locking and loosening is pneumatically actuated. All these
components are vertically aligned. In order to transfer the load to the windshield a head im-
pactor (right-hand side of Figure 1) was pushed against the glass powered by the pneumatic
piston rod. The windshield was put on the four support pins in such a way that the windshield's
central point was positioned under the loading pin. The support pins carry on their upper ends
half-spheres out of Teflon with a diameter of 36 mm. The cylinder is operated in such a way
that the piston before test was locked at about 3/4 of the maximal stroke. After filling the up-
per cylinder chamber with air of about 6 bar pressure and loosening the piston, the piston rod
accelerates up to a constant velocity of about 10 m/s.
The force was measured using a very precise load transducer of HBM, Type S9, load
capacity 5 kN, class 0.05 (measuring errors smaller than 0.05 percent of load capacity). The
displacement of the impactor was measured using a potentiometric displacement sensor of
Buster, Type 8719. For the direct measurement of the acceleration during the impact of the
impactor, an acceleration sensor type of Kistler 8702B500 was used. In addition, the course of
the experiment was analysed with a high-resolution high-speed camera ARAMIS the GOM
mbH for optical 3D deformation analysis. To investigate the influence of the weight of the
windshield, the complete experimental set-up was placed on its side so that the impactor hits
the vertical positioned windshield in the horizontal direction.
Figure 1. Photography of the experimental set-up with the testing frame (left-hand side) and the head im-
pactor (right-hand side).
In cases where the optical identification of the fracture origins in glass is difficult, e.g.
in case of heavy fragmentation, an alternative method of fracture localization is valuable. For
this purpose AE measurements were performed [5]. The AE method utilizes elastic waves,
ÜÜ Four hemispherical
Ü̇̇
ÜÜ
Ü̇̇
3
which were emitted during fracture formation to localize the fracture origin. For the AE meas-
urements the eight channels AMSY 6 system of Vallen was used.
2.2 Wave Velocity Measurement
For the localization of AE sources the wave velocity in the material must be known. Therefore
two sensors (type Vallen VS 900-M) are fixed on the material and an acoustic signal is gener-
ated on a defined position with an exactly known distance to the sensors.
The easiest way to figure out the wave velocity is the “pulsing-function” of the AE
system. After fixing transmitter and receiver in a known distance on surface of the glass plate,
the AE system generates high-voltage wideband pulses. Using the time difference dt between
transmitter and receiver the AE System is able to compute the wave velocity v = s/ dt. Hereby
v is the calculated wave velocity, s the distance between the sensors and dt = t1-t0 the time
difference (see Figure 2).
Figure 2. Schematic of experimental set-up to determine the elastic wave velocity.
In order to determine the influence of the intermediate thin film wave propagation measure-
ments were performed in such a way that a receiver was applied on the upper surface of the
glass plate in distance of s = 770 mm to the transmitter. The transmitter was mounted on the
upper and lower surface of the glass plate, respectively. A high-voltage wideband pulse like a
step function was applied to the transmitter.
Left-hand side of Figure 3 shows the wideband transmitted signal (top) and the detect-
ed signals at the upper (middle) und the lower (bottom) surface of the glass plate, respectively,
and the corresponding frequency spectra (right-hand side of Figure 3). The signals have a
complex waveform, resulting from the superposition of many reflected waves at different an-
gles. There are several types of waves (plate modes) excited with varies wave velocities (so-
called dispersion). The mean velocity of the first mode (corresponds to the fastest extensional
S0 wave) is about 5.2 mm/s, which is slightly below the speed of the longitudinal wave in
glass of about 5.5 mm/s.
Figure 3 clearly shows that the signal waveforms and frequency spectra are very simi-
lar. An influence of the intermediate film to the wave propagation, and thus the location accu-
racy can be excluded. It is not possible to distinguish whether crack propagation starts from
the lower or upper glass plate.
2.3 Localization Test
Another possibility to generate an acoustic emission source is the breakage of a lead of a pen-
cil at the glass plate like shown in Figure 4. This artificial source generation is also known as
Hsu-Nielson source [6]. The so-called pen test simulating a vertical single force drop gener-
ates a signal like a step function with a rise time less than 1 µsec.
The breakage of the brittle lead pencil under an angle of 30° at the top of the glass
plate causes a small elastic deformation of the surface and the surface relaxes immediately,
which induces an elastic wave into the plate. Here it is necessary to use three or more sensors
for a planar localization of the artificial AE source. The time differences of the signals are
needed to calculate the location of the source.
4
Figure 3. Measured waveforms (left-hand side) at the upper (middle) und the lower (bottom) surface of the
glass plate and the corresponding frequency spectra (right-hand side).
In Figure 4 the signals detected at four sensors are plotted. The arrival times of the first
onsets are used to calculate the time differences between each hit sensor. Within these time
differences the localization software generates a hyperbola between each sensor (see Figure 5).
The point of intersection of these hyperbolas is the localized position of the artificial source.
Figure 5 shows the accuracy of the used localization algorithm. The set-up for measuring AE
was as mentioned before. The acoustic wave velocity used for localization was 5.2 mm/s. The
mean deviation between the AE source location and "true" Hsu-Nielson source origin is below
3 mm.
2.4 Test Results
In the following, only the results of one experiment in convex position (impact of outside the
car) and in horizontal direction of loading are described (without the influence of self-weight).
The force-deflection diagram (Figure 6) of this test shows an approximately linear increase of
the force versus deflection. The maximal force of the windshield is about 1500 N and the de-
flection at fracture of 34 mm. After fracturing the windshield shows a residual force of about
400 N.
Figure 7 displays the waveforms of the six AE sensors, which were fixed at the up-
per surface of the windshield. The sequence of the signals corresponds to the sequence of
the measured time onsets. It can be seen that all signals show clearly discernible onsets,
which could automatically determined using a threshold of 40 dB. The many later onsets of
bursts were originated from the formation of the crack meshwork in the glass panel during
the very fast crack propagation.
5
Figure 4. Measured AE signals of an artificial AE source using four sensors at the glass plate. Using the time
differences between the first hit sensor and the others it is possible to locate the AE source.
Figure 5. Localization of the Hsu-Nielson source on a flat plate using four AE sensors. The parabolas are
calculated by differences in the time due of the signal. The small crosses are indicating the position of the four
AE sensors. The grid lines represent a 10-mm spacing.
6
Figure 6. Force versus deflection of a static windshield test in convex position.
Figure 7. Measured AE signals at the upper surface of the AE sensors.
Fo
rce
[N
]
Deflection [mm]
7
Figures 8 and 9 show the projection on the x-y plane with marked contour of the
windshield (Figure 9). The numbers indicate the positions of the six AE sensors at the edge
of the windshield. The location of the AE event is marked by a green dot. The intersection
of the hyperbolas is at the bottom of the windshield at coordinates x = -13.55 cm and y = -
7.02 cm, which corresponds to the location of the AE event. The origin of the coordinate
system is on the vertical line of symmetry 10 cm from the lower edge of the glass pane.
According to the AE measurements the crack starts at the lower edge of the pane.
Figure 8. Projection onto the x-y plane with location of AE event at the intersection of the hyperbolas.
Figure 9. Projection onto the x-y plane with contour of the windshield, the location of AE event (green dot),
and the position of the AE sensors (numbers).
2.4 Verification of AE Results Using Optical Measurement Technique
Displacement and strain measurements were carried out with the optical measuring system
ARAMIS during the experiment to verify the detection results of the AE measurements. The
advantage of this measuring system compared to e.g. measurements with strain gauges is that
a large component can be measured without contact and with a high temporal and spatial reso-
lution. Figure 10 shows at the left-hand side the projection onto the x-y plane with the position
of the AE sensors (compare with Figure 9), the contour of the disk, the location of the AE
event, which corresponds to the initial crack (green dot), and the calculated strain using AR-
AMIS.
5
6
1
2 3
4
8
The right-hand side of Figure 10 shows a photograph of a detail of the windshield after
the experiment. The location of the initial crack is obtained via the optically determined max-
imum strain, which is collared in the figure. In the red-collared area the elongation at fracture
of glass is exceeded. Starting from this point the crack propagates in direction of the centre of
the pane (blue arrow in the right-hand side of Figure 10) where the actual impact has occurred.
The results of the two measurement methods agree very well with the real location of the ini-
tial crack. Although the crack is located at the edge of sensor array the location accuracy is in
the range of a few millimetres.
Figure 10. Left-hand side: Projection onto the x-y plane with calculated strain (left-hand side), contour of the
windshield, and the location of AE event (green dot). Right-hand side: Photography of a detail of the wind-
shield after the test. The blue arrow indicates the initial crack location and the direction of crack propagation.
3 Conclusion
In this article acoustic emission measurements were carried out to locate the initial crack in
laminated safety glass under quasi-static load. The localization results were compared with an
optical measurement method. The crack started from the edge of the windshield and propa-
gates in the direction of the centre of the windshield to the point of the impact. In the edge
region the strength is much lower than in the interior of the plane. This fact can be explained
that grinding of the edges creates small microcracks, which lead to a decrease of the strength.
In accordance with DIN 18008-1 it can be assumed that under tensile load the strength de-
creases up to 20 percent.
In conclusion AE is a helpful tool for the localization of the initial cracks in wind-
shields under impact load. In particular, the initial crack represents an important parameter for
validation and verification of finite element models. A comparison in terms of force versus
deflection shows that the behaviour of laminated glass up to failure is sufficiently repro-
duced. Further investigations must be performed for the simulation of the post breakage
behaviour. Therefore high speed testing of the PVB interlayer and the usage of an adequate
material model are also topics of further investigations.
References
[1] M. Timmel, S. Kolling, P. Osterrieder, P.A. Du Bois: A finite element model for impact simulation with
laminated glass, International Journal of Impact Engineering, 34(8): 1465-1478, 2007.
Risses.
ÜÜ
Localized initial
Ü
9
[2] C. Alter, G. Manthei, S. Kolling, J. Schneider: Localization of initial cracks in laminated glass in the con-
text of large deformations. Proceedings of the international conference at glasstec, Düsseldorf, 2012.
[3] C. Alter, S. Kolling, J. Schneider: Head Impact on Windscreens - Modeling, Validation and Verification.
Proceedings of the 11th
LS-DYNA Forum, Ulm, Germany, 2012.
[4] T. Frank, A. Kurz, M. Pitzer, M. Söllner: Development and validation of numerical pedestrian impactor
models. Proceedings of the 4th
European LS-DYNA Users Conference, pp. C-II-01/18, 2003.
[5] C. U. Grosse, M. Ohtsu (Eds.): Acoustic Emission Testing, Springer, Berlin, Heidelberg, 2008.
[6] N.N. Hsu, J.A. Simmons, S.C.Hardy, An approach to acoustic emission signal analysis – theory and ex-
periment, Materials Evaluation: 100-106, 1977.
Acknowledgement
This research project of the IGF named 17N1111 dealing with ”Experimentelle und numer-
ische Untersuchungen von Windschutzscheiben unter stoßartiger Belastung zur
Verbesserung des Fußgänger- und Insassenschutzes” is supported by the AiF. This was
possible within the framework of the programme to expedite the industrial collective re-
search and development (IGF) of the Federal Ministry for Economic Affairs and Technolo-
gy by a decision of the German Parliament. We would like to thank them for this sponsor-
ship. The work is carried out in co-operation with “Forschungsvereinigung Automobiltech-
nik e.V. (FAT)”.